The Same AI Flooding Your Docket Could Be Running Your Firm

There is a new kind of opposing party showing up in civil litigation, and most firms have not named it yet.

A study out of MIT and the University of Southern California, analyzing 4.5 million federal civil cases and 46 million PACER docket entries, found that AI-generated text in complaints filed by self-represented litigants went from near zero in 2022 to roughly one in five by early 2026. Over the same period, the share of federal civil cases filed without a lawyer climbed from a steady 11 percent, where it had sat for two decades, to almost 17 percent in 2025. That single year saw more than 41,000 pro se filings, nearly double the pre-AI average.

The researchers were direct about the mechanism. Large language models lowered the cost of producing a procedurally viable legal document. The surge clusters exactly where you would expect it to: civil rights complaints, consumer credit disputes, foreclosures. Case types where formulaic drafting does most of the work. Areas that demand sustained specialized knowledge, like patent or securities law, show no measurable effect. AI did not make people better lawyers. It made the paperwork cheap.

The Part That Lands on the Billing Partner's Desk

Here is where this stops being an interesting trend and starts being a line item.

Attorneys interviewed for the research describe client bills climbing from a few thousand dollars into the tens of thousands, driven by the need to read and answer a torrent of AI-generated motions. Many of those motions are repetitive. Some are baseless. A few contain fabricated citations that have to be caught before they cause a problem. None of that matters to the docket, because every filing has to be taken seriously, reviewed, and answered. The volume is the burden, whether or not the substance holds up.

One of the researchers put the system-level version of it plainly: any system that lowers the cost of entry should expect increased demand. The cost of entry to civil litigation just dropped, and the demand is arriving on your associates' desks.

This is not an argument that self-represented litigants should not have access to the courts. They have had that right since the First Congress, and for many people AI is the first realistic path to using it. At least one federal judge has said that AI-assisted filings are often clearer and easier to rule on than the handwritten ones that came before. The access-to-justice story is real, and it is mostly a good one.

But access to justice for the filer and operational load for the firm on the other side are the same event seen from two directions. Your firm is on the receiving end. The question is what you do about the load.

The Uncomfortable Symmetry

The litigants on the other side of your cases have already figured out something a lot of firms are still resisting: AI lowers the cost of legal work. They are using it to generate volume, and the volume is what is straining your margins.

For a self-represented plaintiff, AI lowers the cost of producing one more filing. For a firm, the same class of tools lowers the cost of the work that quietly consumes your week and never shows up as billable time: intake and qualification, drafting first-pass demand letters, keeping clients informed so they stop calling to ask for status, organizing records, and triaging which matters actually deserve partner attention. That is the work that determines how many cases your existing team can carry. It is also the work most firms have never systematized, because until recently the only way to do more of it was to hire another person.

That is the real shift the study points to, even though it never says it directly. The cost structure of legal work is changing on both sides of the v. Firms that build their own AI layer absorb the rising volume without adding headcount. Firms that do not will feel it the way the lawyers in the study already do: as bills that balloon, associates who burn out reading junk motions, and a practice that gets more expensive to run every quarter.

The Tool Is Already in the Courtroom

You do not get to vote on whether AI is part of litigation now. The MIT and USC data settles that. It is here, it is on the other side of your cases, and its share of the docket is rising every year.

The only decision still open to you is which side of your own operation it works on. Either it is a force multiplier for the people suing your clients and a cost you eat, or it is the leverage that lets your firm handle more matters, serve clients faster, and keep your margins intact while the volume climbs.